Research map / Probabilistic and scientific modeling / Variational and generative sampling
Deconvolutional Latent-Variable Model for Text Sequence Matching
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use a deconvolutional rather than LSTM sequence decoder | LVM · BookCorpus/arXiv style classification | accuracy: 91.7 -> 96.2 | Helped |
| Use a deconvolutional rather than LSTM sequence decoder | LVM · Unsupervised training, 10,000 iterations | training time: 39m 41s -> 8m 23s | Helped |
| Use a deconvolutional rather than LSTM sequence decoder | LVM · SNLI, semi-supervised | accuracy: 64.7 -> 67.2; 67.5 -> 69.3; 71.1 -> 72.2 | Helped |
| Use a deconvolutional rather than LSTM sequence decoder | LVM · Quora Question Pairs | accuracy: 62.9 -> 65.1; 67.6 -> 69.4; 69.0 -> 70.5; 72.4 -> 73.7 | Helped |
| Add the variational latent-variable objective to the deconvolutional autoencoder (ablation) | DeConv-AE · SNLI, semi-supervised | accuracy: 62.1 -> 67.2; 65.5 -> 69.3; 68.7 -> 72.2 | Helped |
| Add the variational latent-variable objective to the deconvolutional autoencoder (ablation) | DeConv-AE · Quora Question Pairs | accuracy: 60.2 -> 65.1; 65.1 -> 69.4; 67.7 -> 70.5; 71.6 -> 73.7 | Helped |
| Use the deconvolutional latent-variable model instead of the deconvolutional autoencoder (ablation) | DeConv-AE · SNLI, all labeled data | accuracy: 81.6% -> 80.9% | Hurt |
| Use a deconvolutional rather than LSTM decoder | LVM · Unsupervised training | number of parameters: ~ 16 million -> ~ 12 million | Helped |
Closest papers by meaning
- Deconvolutional Paragraph Representation Learning (NeurIPS 2017 · similarity 0.89)
- Latent Sequence Decompositions (ICLR 2017 · similarity 0.88)
- A Compare-Aggregate Model for Matching Text Sequences (ICLR 2017 · similarity 0.87)
- DenoSent: A Denoising Objective for Self-Supervised Sentence Representation Learning (AAAI 2024 · similarity 0.84)
- Language Model Decoding as Direct Metrics Optimization (ICLR 2024 · similarity 0.84)
- Repurposing Language Models into Embedding Models: Finding the Compute-Optimal Recipe (NeurIPS 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.